Removed web/main/svn, svn-web symlinks because they need to be .htaccess-es in order for the relative mod_rewrite commands to work correctly

Added web/main/svn, svn-web symlinks to Redmine/* for shorter URLs

Added web/main/Redmine/svn-web/

inputs/GBIF/: Added scripts for subsetting refresh

lib/sql.py: table_order_by(): Documented that it returns None if table is a view, because table_cluster_on() would return None. This is necessary for inputs/FIA/occurrence_all/ sorting to work correctly, because specifying a manual sort order would prevent the query planner from just using fast nested loop joins and instead cause it to perform a slow sort. (This appears to be a bug in the query planner, because when the column list specified matches the joined-on indexes, there should be no need for post-nested loop re-sorting.)

inputs/FIA/occurrence_all/test.xml.ref: Updated inserted row count for new row sort order

lib/db_xml.py: put_table(): Fixed bug where also need to advance start to fetch next set when table is a view, because the views that are now being used with the import (inputs/FIA/occurrence_all/) are static rather than dynamic and do not return different rows after the previous set of rows has been imported

inputs/FIA/occurrence_all/import: Removed no longer applicable comment that directional joins are needed for PostgreSQL query planner to avoid slow sorts

inputs/FIA/TREE/import: Reclustered table by TREE.parent path index, to facilitate path-order joins

inputs/FIA/occurrence_all/import: Changed all RIGHT JOINs to inner joins so that tables would be joined in path order (i.e. general->specific). This optimizes the incremental joins so that the small tables are joined to each other before being joined to the large tables, rather than each row of the large tables being looked up in the small tables. This effect may not be noticeable for small LIMIT values, but would become apparent for large LIMIT values, such as the 1-million-row partitions used by db_xml.put_table() for column-based import. Note that inner joins used to cause the query planner to produce incorrect results containing slow sorts, but now this appears to no longer be an issue, perhaps because the result is not sorted by the TREE.ID index (which is not in the same order as the path indexes *.unique, *.parent).